ETISEO%20Project

1
ETISEO Project
Video corpus dedicated to the first evaluation
2
ETISEO Project

• To capture realistic video sequences with
  graduate difficulties
• lighting variations,
• occlusions
• covering predefined scenarios.

3
ETISEO Project
Video Providers Corpus Data description Video
contents demonstration
4
Video providers
Several data providers offering large diversity.

• Different sites
• corridor,
• sub-way,
• - building entry,
• - apron,
• - car park.

• Different contexts
• indoor / outdoor
• single / multi-camera
• big / small overlap area
• visible / IR
• actors / real scenes

5
Video providers Silogic

• Toulouse-Blagnac Airport
• France

Apron
6
Video providers Silogic
Apron
Multi-view
7
Video providers INRETS
Building Entry
Multi-view
Car Park
8
Video providers CEA
Corridor
Car street
9
Video providers CEA
Geometry IR/Colour calibration ?Estimation of
an homographic matrix between the 2 images
(sensors close and aligned)
10
Video providers RATP
Subway
Real scene
11
Corpus data
Video Providers Corpus Data description Video
contents demonstration
12
Priority Sequences

• Evaluation risk

• That participants process only part of the
  dataset.

• Resulting in dispersed metrics.

• No real evaluation or comparison being possible

13
Priority Sequences

• Solution

• Create some mandatory sequences.

• Participants should process at least these
  sequences.

• Need some feedback to select these priority
  sequences. See the questionnaire.

14
Corpus data
Provider Site Sequence Persons Vehicle Multiview
Silogic Apron 1 to 6 ? ? ?
RATP Station 7 to 10 ?
CEA Corridor 11 to 13 ? IR / visible
CEA Road 14 to 17 ? ? IR / visible
INRETS Building entry 18 to 20 ? ? ?
For each topic, the set contains sequences with
graduate difficulty.
15
Corpus data

• Covered scenarios
• vehicle parking
• people getting in out from a vehicle
• people walking
• people crossing
• person entering a building
• person entering a room
• person meeting
• abandoned baggage

16
Corpus data

• Summary
• indoor / outdoor sequences,
• 8 scenarios,
• graduate difficulty,
• 20 sequences,
• 37 video clips,
• 5.2 Go of data.

17
Corpus diffusion

• Video format
• sequence of jpeg/bmp images for colour images
• sequence of TIFF and jpeg images for IR.
• Timestamp
• Image name contains the timestamp relative to the
  start of the sequence.
• Calibration
• Provide as
• - a calibration matrix used to generate 3D
• - couples of 2D / 3D points.

18
Corpus diffusion

• Structure

Contains the sequence
ETI-VS2-RD-32
ETI-VS2-RD-32-C1
Camera folders
C1-Format.xml
ETI-VS2-RD-32-C1-00h00m36s000-0000.jpeg
ETI-VS2-RD-32-C1-.jpeg
ETI-VS2-RD-32-C1-00h02m40s719-1559.jpeg
ETI-VS2-RD-32-C2
C2-Format.xml
ETI-VS2-RD-32-C2-00h00m36s000-0000.jpeg
ETI-VS2-RD-32-C2-.jpeg
ETI-VS2-RD-32-C2-00h02m40s723-1559.jpeg
19
Corpus diffusion

• Structure
• Each video resource will be store in a unique
  folder using the following convention
•  ETI-VSindexdataset-Type-sequenceindex
• Example
• ETI-VS2-RD-32

Type of Video Acronym used
Parking PK
Road RD
Apron AP
Metro MO
Building entrance BE
Building corridor BC
20
Corpus diffusion

• C1-format.xml
• lt?xml version"1.0" encoding"UTF-8"?gt
• ltsequence nameETI-VS1-BC-32gt
• ltpropertiesgt
• ltformatgtJPEGlt/formatgt could be
  JPEG,TIFF
• ltwidthgt720lt/widthgt
• ltheightgt576lt/heightgt
• ltdepthgt24lt/depthgt
• ltframesgt1560lt/framesgt of frames
  (start0)
• ltinter-frame-msgt80000lt/inter-frame-msgt average
  time in ms between frames
• ltmaxsizegt65232lt/maxsizegt maximum image
  size
• ltcameragtC1lt/cameragt camera id
• ltsensorgtcolorlt/sensorgt color, IR,
  greyscale
• lt/propertiesgt
• lt/sequencegt

21
ETISEO Project
Video Providers Corpus Data description Video
content demonstration
22
Video demonstration

• Silogic / Apron
• Type multi-view, played and real scenes.
• Precision evaluation scenes with real 3D ground
  truth.
• person walking on the zone. C1, C2, C3, C4
• vehicle moving on the zone. C1, C2, C3, C4
• Real scenes vehicles moving on the apron with
  large variation of conditions (outside
  environment).
• GPU arrival  normal conditions . C1 , C2
• GPU arrival   reflections . C1 , C2
• GPU arrival   shadows  C1 , C2

23
Video demonstration

• INRETS / parking building entry.
• Type multi-view, played scenes.
• Scenes vehicles parking, people
  entering/leaving cars, entering/leaving the
  building. Large variation of conditions (outside
  environment).
• ETI-VS1-BE-18 C1,C2,C3,C4
• ETI-VS1-BE-19 C4
• ETI-VS1-BE-20 C1,C2,C3,C4

24
Video demonstration

• INRETS / parking building entry.
• Type multi-view, played scenes.
• Scenes Sequence ETI-VS1-BE-20 will be provided
  with different compression levels
• BMP, JPEG and MPEG.

25
Video demonstration

• CEA / parking building entry.
• Type single-view but visible/IR couple. Played
  scenes.
• Road vehicles and people.
• ETI-VS1-RD-14 C1 , C2
• ETI-VS1-RD-15 C1 , C2
• ETI-VS1-RD-16 C1 , C2
• ETI-VS1-RD-17 C1 , C2
• Corridor People crossing, abandoned luggage.
• ETI-VS1-BC-11 C1 , C2
• ETI-VS1-BC-12 C1 , C2
• ETI-VS1-BC-13 C1 , C2

26
Video demonstration

• RATP / subway.
• Type single-view. Played scenes mixed with real
  environment.
• Corridor People crossing, abandoned luggage.
• ETI-VS1-MO-07
• ETI-VS1-MO-08
• ETI-VS1-MO-09
• Railway platform People crossing, abandoned
  luggage.
• ETI-VS1-MO-10